Clearview AI has built and tested an unreleased prototype called InquiryIQ that uses generative AI to automatically assemble profiles of people identified through its face-recognition system, with one tested model coming from SpaceXAI, the entity formed by the February merger of SpaceX and Grok-maker xAI. The tool is designed to take a lead from a face-recognition hit and then crawl the open web to surface aliases, employers, addresses, phone numbers, associates, arrest history, and social media accounts. Clearview's database has grown from more than 3 billion scraped photos in 2020 to over 70 billion today, and the company says more than 2,000 law enforcement agencies use its technology. Details of the prototype were surfaced by WIRED, which reconstructed the tool from code exposed by Clearview's login page.
InquiryIQ's interface tells investigators that supplying age, gender, and race helps the system make "smarter decisions" as it searches, according to the code reviewed. The tool is designed to run web and image searches, browse pages, apply face recognition to photos it encounters, and build what Clearview calls a "Candidate Graph" of possible identities and connections. The system flags automatically generated demographic, social, and arrest data as potentially inaccurate, and officers must attest they independently verified any finding before it enters the profile.
Clearview disputes the framing that InquiryIQ is an automated investigator. CEO Amos Kyler, who joined the company as an engineer in 2019 and took over as CEO in October 2025, describes it as a limited way to automate the web searches detectives already perform.
“No law enforcement user has ever used it, period.”— Amos Kyler, Clearview AI CEO
Key facts
- 01Clearview AI's face-recognition database has grown from 3 billion images in 2020 to over 70 billion today.
- 02The company says its tools are used by more than 2,000 law enforcement agencies nationwide.
- 03InquiryIQ tested a model from SpaceXAI, the entity formed by the February merger of SpaceX and xAI.
- 04CEO Amos Kyler says no law enforcement user has ever used InquiryIQ and it is not planned for release in its current form.
- 05Peter Thiel invested $200,000 in Clearview in 2017, the year the company was founded.
Kyler says the SpaceXAI option and other models were surfaced in the interface only so Clearview's engineers could compare performance during testing, not so police could pick which model ran the search. He frames the tool's logic as hypothesis-testing: taking a "factoid" an investigator has flagged and asking whether the wider web supports or contradicts it. Clearview says modern AI tooling has made it faster to build sophisticated prototypes, and the company's internal directive is to push those prototypes along quickly.
The choice of Grok's underlying model is notable given the chatbot's track record of racist and extremist outputs since Elon Musk positioned it as an alternative to what he calls "woke" AI systems. Clearview declined to speculate on how SpaceXAI would use the demographic inputs InquiryIQ collects or how those inputs would shape outputs. Other intelligence platforms sold to law enforcement, including ShadowDragon's SocialNet, Penlink's Tangles, and Fivecast, already automate parts of this work, and Flock Safety was recently found to be developing a similar AI tool called OS Investigate.
Andrew Guthrie Ferguson, a George Washington University law professor who studies AI and policing, calls this class of tool "digital rummaging" — gathering scattered digital breadcrumbs that once lived in separate places and were difficult to connect.
The practical concern raised by legal scholars is that compressing days of detective work into minutes lowers the cost of fishing expeditions. Woodrow Hartzog, a Boston University privacy scholar, argues that the labor of investigation historically served as a check on state surveillance. Rules against dragnet-style monitoring were never fully written because the effort required to follow everyone made those rules unnecessary. Generative AI removes that friction, and it can also produce different answers from identical starting points, making it hard to reconstruct why a system chased one lead over another.
Clearview points to its human-in-the-loop requirement as the safeguard. Kyler says an analyst's job is to sort signal from noise and decide what is real.
“A human in the loop is a little bit of a cold comfort.”— Woodrow Hartzog, Boston University privacy scholar
Hartzog is skeptical of that safeguard in practice, noting that investigators tend to defer to automated outputs over time until human review becomes ceremonial rather than substantive. That pattern — automation bias creeping into workflows where the machine's output is presumed correct — has been documented across other predictive policing and forensic AI systems.
Clearview's arc has been unusually loud for a surveillance company. Founded in 2017 with a $200,000 investment from Peter Thiel, it operated quietly until The New York Times revealed in 2020 that it had scraped more than 3 billion images from Facebook, YouTube, Venmo, and millions of other sites. Founder Hoan Ton-That stepped down as CEO in December 2024 and left the board in spring 2025 after years of controversy over his past ties to far-right figures, for which he apologized. Kyler now runs the company and emphasizes controls, auditing, and refinement over Clearview's earlier boundary-pushing posture.
The prototype's existence matters even if it never ships in its current form. The market for AI-assisted open-source intelligence sold to police is clearly moving from keyword search and link analysis toward autonomous agents that construct dossiers on demand. Once the underlying models are cheap enough and the interfaces are polished, the operational question for every US law enforcement agency shifts from whether to run such a search to how often. Clearview, sitting on 70 billion images and 2,000 agency relationships, is positioned to be the default distribution channel — and the regulatory frameworks built for a slower world will decide the terms on which that distribution happens.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
